ARCANE-PedSynth: Synthetic Multi-Pedestrian Datasets with Behavioural Crossing Annotations
Researchers have developed ARCANE-PedSynth, an open-source framework built on CARLA for generating synthetic datasets of pedestrians. This framework uses a hybrid AI-manual control system to achieve significantly higher pedestrian crossing rates than CARLA's default. It produces synchronized sensor data like RGB and LiDAR, along with detailed behavioral annotations and pose keypoints, demonstrated through the PedSynth++ dataset. AI
IMPACT Enables more robust training of autonomous driving systems by providing diverse synthetic pedestrian crossing scenarios.